View in EDS HTML Full Text PDF Full Text

When Tutors Simultaneously Instruct Students from the Primary, Middle, and High School Levels in Online One-on-One Tutoring: Investigating the Interaction Dynamics Using AI, ENA, and LSA Methods

Saved in:
Bibliographic Details
Title: When Tutors Simultaneously Instruct Students from the Primary, Middle, and High School Levels in Online One-on-One Tutoring: Investigating the Interaction Dynamics Using AI, ENA, and LSA Methods
Language: English
Authors: Deliang Wang (ORCID 0009-0008-6488-0234), Lei Gao, Dapeng Shan, Gaowei Chen, Chenwei Zhang, Ben Kao
Source: Journal of Science Education and Technology. 2025 34(5):1128-1142.
Availability: Springer. Available from: Springer Nature. One New York Plaza, Suite 4600, New York, NY 10004. Tel: 800-777-4643; Tel: 212-460-1500; Fax: 212-460-1700; e-mail: customerservice@springernature.com; Web site: https://link.springer.com/
Peer Reviewed: Y
Page Count: 15
Publication Date: 2025
Document Type: Journal Articles
Reports - Research
Education Level: Elementary Education
Junior High Schools
Middle Schools
Secondary Education
High Schools
Descriptors: Tutors, Tutoring, Peer Teaching, Computer Mediated Communication, Interaction, Elementary School Students, Middle School Students, High School Students, Mathematics Instruction, Teaching Methods, Tutor Training, Teaching Skills, Dialogs (Language), Instructional Effectiveness, Efficiency
DOI: 10.1007/s10956-024-10154-4
ISSN: 1059-0145
1573-1839
Abstract: Online one-on-one tutoring serves as a personalized approach to supplement classroom instruction. However, with the growing tutoring market, a single tutor often handles inquiries from students across primary, middle, and high school levels. Consequently, the extent of tutors' interactions with students of varying grades and their use of tutoring strategies to enhance student learning remains unclear. To address this gap, we collected and analyzed 1500 tutoring dialogues from amateur mathematics tutors concurrently instructing students from primary, middle, and high school levels. These dialogues were annotated using a coding scheme and a well-trained powerful artificial intelligence (AI) model. The interaction dynamics were subsequently examined using epistemic network analysis and lag sequential analysis, yielding findings on the occurrences, co-occurrences, and sequential patterns of dialogic strategies. First, the results reveal that tutors frequently engaged in off-topic behaviors and direct instruction, regardless of students' educational level. Second, tutors' constructive questions and knowledge sharing and instruction were more associated with greater constructive expressions from students at higher educational levels, while primary students primarily demonstrated simple acknowledgment. Third, tutors exhibited limited sequential patterns of dialogic strategies when tutoring primary and middle school students, mainly focusing on question-asking behaviors and evaluation and feedback. In contrast, tutors displayed diverse patterns across various categories of dialogic strategies when instructing high school students, emphasizing the facilitation of students' reasoning and metacognition. These findings underscore the importance of training tutors to develop dialogic skills and adopt tailored pedagogical approaches for different educational levels, ensuring effective and efficient online one-on-one mathematics tutoring.
Abstractor: As Provided
Entry Date: 2026
Accession Number: EJ1497295
Database: ERIC
Full text is not displayed to guests.
FullText Links:
  – Type: pdflink
    Url: https://content.ebscohost.com/cds/retrieve?content=AQICAHj0k_4E0hTGH8RJwT4gCJyBsGNe_WN95AvKlDbXJGqwxwGuNNxDcZ6DGUn2I0ouLA7mAAAA4zCB4AYJKoZIhvcNAQcGoIHSMIHPAgEAMIHJBgkqhkiG9w0BBwEwHgYJYIZIAWUDBAEuMBEEDFVNjlySqasecDF6HQIBEICBm2oI8WBFJjPancXYmr1mvoLSnztyiPYEy5GgbH46qKOQcbfBafAM5fag5NZyouNblHx3Oy12Bwzf2fyIOFyXmwUKi1eelmqCj3nBaqTdFq-n4GmYVGo5rauhGm-eeyVc5WcxfSZ7RSAp7IBBUwok0yuD_4jZsqjfJnQ0dWuJ_g0xtzbBw_5M0yEjUeE5aauPkP5nhqlFbYihHtqh
Text:
  Availability: 1
  Value: <anid>AN0189590664;4n601oct.25;2025Nov28.05:32;v2.2.500</anid> <title id="AN0189590664-1">When Tutors Simultaneously Instruct Students from the Primary, Middle, and High School Levels in Online One-on-One Tutoring: Investigating the Interaction Dynamics Using AI, ENA, and LSA Methods </title> <p>Online one-on-one tutoring serves as a personalized approach to supplement classroom instruction. However, with the growing tutoring market, a single tutor often handles inquiries from students across primary, middle, and high school levels. Consequently, the extent of tutors' interactions with students of varying grades and their use of tutoring strategies to enhance student learning remains unclear. To address this gap, we collected and analyzed 1500 tutoring dialogues from amateur mathematics tutors concurrently instructing students from primary, middle, and high school levels. These dialogues were annotated using a coding scheme and a well-trained powerful artificial intelligence (AI) model. The interaction dynamics were subsequently examined using epistemic network analysis and lag sequential analysis, yielding findings on the occurrences, co-occurrences, and sequential patterns of dialogic strategies. First, the results reveal that tutors frequently engaged in off-topic behaviors and direct instruction, regardless of students' educational level. Second, tutors' constructive questions and knowledge sharing and instruction were more associated with greater constructive expressions from students at higher educational levels, while primary students primarily demonstrated simple acknowledgment. Third, tutors exhibited limited sequential patterns of dialogic strategies when tutoring primary and middle school students, mainly focusing on question-asking behaviors and evaluation and feedback. In contrast, tutors displayed diverse patterns across various categories of dialogic strategies when instructing high school students, emphasizing the facilitation of students' reasoning and metacognition. These findings underscore the importance of training tutors to develop dialogic skills and adopt tailored pedagogical approaches for different educational levels, ensuring effective and efficient online one-on-one mathematics tutoring.</p> <p>Keywords: Online one-on-one tutoring; Mathematics; Tutor–student interaction; Epistemic network analysis; Lag sequential analysis; Artificial intelligence; Education Curriculum and Pedagogy Specialist Studies In Education</p> <p>Copyright comment Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.</p> <hd id="AN0189590664-2">Introduction</hd> <p>Tutoring is defined as an instructional approach where a tutor, possessing sufficient knowledge and skills in a specific subject area, imparts their expertise to a tutee who possesses less knowledge in that area (Graesser et al., [<reflink idref="bib20" id="ref1">20</reflink>]; Hanham et al., [<reflink idref="bib25" id="ref2">25</reflink>]; Wood et al., [<reflink idref="bib74" id="ref3">74</reflink>]). Recognized as a supplement to traditional classroom instruction, tutoring has demonstrated its ability to enhance student learning outcomes over the past few decades (Bloom, [<reflink idref="bib3" id="ref4">3</reflink>]; De Ree et al., [<reflink idref="bib12" id="ref5">12</reflink>]; Fuchs et al., [<reflink idref="bib17" id="ref6">17</reflink>]; Hardt et al., [<reflink idref="bib26" id="ref7">26</reflink>]; Hume et al., [<reflink idref="bib31" id="ref8">31</reflink>]; Lin et al., [<reflink idref="bib37" id="ref9">37</reflink>]). In a systematic review and meta-analysis conducted by Nickow et al. ([<reflink idref="bib49" id="ref10">49</reflink>]), which examined 96 PreK12 tutoring programs, tutoring has been found to have consistent and substantial positive impacts on academic performance, with an overall pooled effect size estimate of 0.37 standard deviations, particularly in the subjects of mathematics and reading. Compared to traditional classroom instruction, tutoring provides a higher level of flexibility in duration and individualized support tailored to the unique needs of students (Robinette, [<reflink idref="bib54" id="ref11">54</reflink>]; Št'astny', [<reflink idref="bib62" id="ref12">62</reflink>]).</p> <p>The advancement of the internet and technology has brought about significant changes in the field of tutoring. One notable development is the emergence of online tutoring, which transcends temporal and spatial limitations (Zhang et al., [<reflink idref="bib77" id="ref13">77</reflink>]). Its prominence has further increased during and after the COVID-19 pandemic (Hardt et al., [<reflink idref="bib26" id="ref14">26</reflink>]). Another noteworthy change is the shift towards a more individualized and flexible tutor selection process, such as the popularity of online one-on-one tutoring. Instead of being limited to a fixed tutor, students now have the ability to choose from a pool of suitable tutors through pairing algorithms implemented in tutoring systems or platforms (e.g., Snapask, [<reflink idref="bib59" id="ref15">59</reflink>]). Some empirical studies have also shown that online one-on-one tutoring offers a superior learning experience, such as improving students' perceived confidence and satisfaction (Teakel et al., [<reflink idref="bib64" id="ref16">64</reflink>]), and has a positive impact on tutees' academic performance in reading and mathematics (e.g., Chappell et al., [<reflink idref="bib6" id="ref17">6</reflink>]; Nickow et al., [<reflink idref="bib49" id="ref18">49</reflink>]; Tsuei, [<reflink idref="bib67" id="ref19">67</reflink>]; Zhang et al., [<reflink idref="bib77" id="ref20">77</reflink>]).</p> <p>The global market for online tutoring has witnessed remarkable growth, with its value reaching USD 7.69 billion in 2022 and its projections indicating a substantial increase to approximately USD 23.73 billion by 2030 (Grand View Research, [<reflink idref="bib23" id="ref21">23</reflink>]). This expansion has created a pressing demand for recruiting more tutors to provide personalized services for students. Typically, online one-on-one tutoring involves a tutor instructing a single student at a specific educational level during a given session. For instance, a tutor may primarily teach high school students. However, due to the limited availability of tutors (Joshi, [<reflink idref="bib33" id="ref22">33</reflink>]) and the high volume of student inquiries, it has been observed that individual tutors may receive and address questions from not only high school students but also primary and middle school students, as evidenced by tutoring records from online tutoring platforms, such as "Snapask." In such scenarios, tutors are unable to well prepare their instructional plans in advance. The effectiveness of this online tutoring structure on students' learning outcomes largely depends on tutors' teaching strategies—whether they just answer students' questions directly or scaffold learning step by step (Bates et al., [<reflink idref="bib2" id="ref23">2</reflink>]; Zhang et al., [<reflink idref="bib77" id="ref24">77</reflink>]). However, research indicates that the primary hiring criteria for tutors typically prioritize strong subject matter knowledge (i.e., surpassing the subject knowledge of the tutees) (Booth et al., [<reflink idref="bib4" id="ref25">4</reflink>]). This results in many tutors entering the field without certified teaching licenses or prior teaching experience (i.e., amateur tutors) (Nickow et al., [<reflink idref="bib49" id="ref26">49</reflink>]). Consequently, tutors may feel challenged in providing effective learning support for students. For example, a qualitative study by Joubert and Snyman ([<reflink idref="bib34" id="ref27">34</reflink>]) found that tutors lacked adequate teaching training and did not know how to handle students with low participation and commitment. All these factors raise concerns about the tutors' ability to effectively engage with students across different grade levels to facilitate optimal student learning outcomes.</p> <p>Despite the importance of this issue, research on the dynamics of student–teacher interactions within the context of online tutoring remains limited (Hanham et al., [<reflink idref="bib25" id="ref28">25</reflink>]; Zhang et al., [<reflink idref="bib77" id="ref29">77</reflink>]), particularly regarding the scenario of one tutor instructing students from multiple school levels in one-on-one tutoring. This research gap hinders our understanding of the characteristics of dialogic interaction in online tutoring and impedes the design and implementation of effective training programs to enhance tutors' professional development (McFarlane, [<reflink idref="bib42" id="ref30">42</reflink>]). To address this gap, this study collects online one-on-one tutoring dialogue from tutors who simultaneously instruct students from primary, middle, and high schools and evaluates whether they can employ diverse tutoring strategies to handle students from different educational levels. The research question directing this study is as follows:</p> <p></p> <ulist> <item> RQ: How do tutors utilize and concurrently employ dialogic strategies in online one-on-one tutoring sessions with students at primary, middle, and high school levels, and what sequential patterns emerge in their interactions?</item> </ulist> <hd id="AN0189590664-3">Literature Review</hd> <p></p> <hd id="AN0189590664-4">Online Tutoring and Dialogic Teaching</hd> <p>Online tutoring is an approach in which a tutee engages in inquiry-based learning activities using internet technologies while receiving subject-specific support from a tutor (Hung, [<reflink idref="bib32" id="ref31">32</reflink>]). When only a single tutor and tutee are involved, it is treated as online one-on-one tutoring. Depending on the initiator of the tutoring session, online one-on-one tutoring can be categorized as tutor-centered or student-centered (Zhang et al., [<reflink idref="bib77" id="ref32">77</reflink>]). In student-centered tutoring, students initiate the session by posing questions and engaging in discussions with tutors, leading to problem-solving and knowledge acquisition. On the other hand, in tutor-centered tutoring, tutors guide students according to a tutorial plan that is well-prepared in advance. Student-centered tutoring offers greater flexibility, allowing students to initiate sessions based on their needs, while tutors only need to respond to students' inquiries without preparing materials beforehand (Zhang et al., [<reflink idref="bib77" id="ref33">77</reflink>]). However, the unpredictable nature of student-centered tutoring presents challenges for tutors. For example, the uncertainty of session duration and students' learning objectives require tutors to adopt appropriate strategies to address these situations.</p> <p>To facilitate effective online tutoring, researchers have identified several essential skills that tutors should possess (Chappell et al., [<reflink idref="bib6" id="ref34">6</reflink>]; Chi et al., [<reflink idref="bib8" id="ref35">8</reflink>]; Hampel & Stickler, [<reflink idref="bib24" id="ref36">24</reflink>]; McPherson & Nunes, [<reflink idref="bib43" id="ref37">43</reflink>]). Among these skills, pedagogical abilities play a vital role. Dialogic teaching, an instructional approach that emphasizes the power of dialogue to enhance student engagement and learning, has been recognized as particularly valuable (Howe & Abedin, [<reflink idref="bib29" id="ref38">29</reflink>]; Kim & Wilkinson, [<reflink idref="bib35" id="ref39">35</reflink>]). In the context of online tutoring, dialogue serves as the primary means of social interaction between tutors and students, occurring either verbally or through text-based communication (Wang et al., [<reflink idref="bib73" id="ref40">73</reflink>]). By employing dialogic strategies, tutors aim to elicit active responses from students, engage in productive conversations, address their queries, and facilitate knowledge construction. However, studies have revealed that professional teachers often struggle to master and apply dialogic skills, as traditional initiation-response-evaluation/follow-up (IRE/F) patterns tend to dominate their instruction (Chen, [<reflink idref="bib7" id="ref41">7</reflink>]; Resnick et al., [<reflink idref="bib53" id="ref42">53</reflink>]). This challenge is further amplified for amateur tutors lacking adequate teaching experience, who may be simultaneously tutoring students from different grade levels. Therefore, investigating the characteristics of dialogic interaction between such tutors and students is necessary to further enhance the effectiveness and efficiency of online tutoring.</p> <hd id="AN0189590664-5">Coding Framework for Analyzing Online Tutoring Dialogue</hd> <p>To enable quantitative analysis of educational dialogue, coding frameworks are usually necessary for annotating the features within the dialogue (e.g., dialogic acts) (Mercer, [<reflink idref="bib44" id="ref43">44</reflink>]; Wang et al., [<reflink idref="bib73" id="ref44">73</reflink>]). Thus, a number of coding frameworks have been developed to quantitatively analyze educational dialogue in different educational contexts and tasks. For instance, there are coding frameworks specifically designed for coding teachers' questions (e.g., Chin, [<reflink idref="bib10" id="ref45">10</reflink>]) and frameworks for analyzing whole-class discussions (e.g., Michaels et al., [<reflink idref="bib46" id="ref46">46</reflink>]). Recent systematic reviews conducted by Hennessy et al. ([<reflink idref="bib27" id="ref47">27</reflink>]), Song et al. ([<reflink idref="bib60" id="ref48">60</reflink>]), and Tao and Chen ([<reflink idref="bib63" id="ref49">63</reflink>]) have examined commonly used coding schemes for educational dialogue. Widely recognized frameworks include the academically productive talk (APT) framework (or accountable talk theory; Michaels et al., [<reflink idref="bib45" id="ref50">45</reflink>]), the interactive, constructive, active, and passive (ICAP) framework (Chi & Wylie, [<reflink idref="bib9" id="ref51">9</reflink>]), the scheme for educational dialog analysis (SEDA) (Hennessy et al., [<reflink idref="bib28" id="ref52">28</reflink>]), and the Cambridge dialogue analysis scheme (CDAS) (Hennessy et al., [<reflink idref="bib27" id="ref53">27</reflink>]). These well-established frameworks, built upon robust theoretical foundations, have also been adapted or extended by researchers in their respective studies (e.g., Chen, [<reflink idref="bib7" id="ref54">7</reflink>]; Vrikki et al., [<reflink idref="bib72" id="ref55">72</reflink>]).</p> <p>In the realm of online tutoring dialogue, various coding frameworks have been developed. For instance, Mitchell et al. ([<reflink idref="bib47" id="ref56">47</reflink>]) and Vail and Boyer ([<reflink idref="bib70" id="ref57">70</reflink>]) designed a two-level annotation scheme specifically tailored for identifying dialogue acts in online tutoring. These acts include hints, feedback, questions, answers, statements, acknowledgements, and confirmations. The effectiveness of this scheme in capturing the online tutoring process has been demonstrated in several studies (Lin et al., [<reflink idref="bib36" id="ref58">36</reflink>]; Lin et al., [<reflink idref="bib37" id="ref59">37</reflink>]; Vail et al., [<reflink idref="bib69" id="ref60">69</reflink>]). Likewise, Litman and Forbes-Riley ([<reflink idref="bib38" id="ref61">38</reflink>]) devised a coding scheme that enables the annotation of tutor and student question acts, tutor feedback acts, tutor state acts, and student answer acts. By determining dialogic acts for each utterance, researchers have explored investigating the relationship between these dialogic strategies and students' learning outcomes (e.g., Boyer et al., [<reflink idref="bib5" id="ref62">5</reflink>]; Vail & Boyer, [<reflink idref="bib70" id="ref63">70</reflink>]). For instance, Boyer et al. ([<reflink idref="bib5" id="ref64">5</reflink>]) discovered that tutors who employed more positive cognitive feedback contributed positively to students' learning gains. In addition to the classification of dialogic acts, coding frameworks have also been developed for other dimensions of online tutoring dialogue. For instance, Xu et al. ([<reflink idref="bib76" id="ref65">76</reflink>]) designed a coding scheme to identify different types of instructions used in online tutoring, including greeting, guidance, note-taking, commending, repeating, and summarization.</p> <hd id="AN0189590664-6">Quantitative Analytical Methods for Educational Dialogue</hd> <p>Upon determining coding schemes, researchers typically annotate educational dialogues using either manual or automatic methods (Mercer, [<reflink idref="bib44" id="ref66">44</reflink>]; Wang et al., [<reflink idref="bib73" id="ref67">73</reflink>]). Manual annotation involves human annotators labeling the data (Mercer, [<reflink idref="bib44" id="ref68">44</reflink>]), while automatic annotation leverages artificial intelligence tools for labeling, gaining popularity among researchers over the past decade due to its ability to efficiently annotate large-scale data (Wang et al., [<reflink idref="bib73" id="ref69">73</reflink>]). Once each utterance or turn in the educational dialogue is annotated with a specific label, researchers commonly employ descriptive statistics, significance tests, epistemic network analysis, and sequential pattern mining techniques for quantitative analysis.</p> <p>Descriptive statistics utilize fundamental descriptions to present findings, such as the average number of teacher and student messages in a tutoring session (Shan et al., [<reflink idref="bib58" id="ref70">58</reflink>]). Significance tests assess whether observed results are statistically meaningful or could have occurred by chance (e.g., whether the number of productive talks in a setting is significantly higher than in other settings; Chen, [<reflink idref="bib7" id="ref71">7</reflink>]).</p> <p>Epistemic Network Analysis (ENA), an emerging quantitative ethnography method, innovatively analyzes the co-occurrences of coded data in educational dialogues (Shaffer, [<reflink idref="bib56" id="ref72">56</reflink>]; Zhang et al., [<reflink idref="bib78" id="ref73">78</reflink>]). Drawing on epistemic frame theory, ENA emphasizes the importance of connections between discourse elements rather than their mere presence (Shaffer & Ruis, [<reflink idref="bib55" id="ref74">55</reflink>]). ENA aims to use networks to visualize connections and strength of associations among different discourse elements (Shaffer & Ruis, [<reflink idref="bib57" id="ref75">57</reflink>]), addressing the limitations of traditional frequency-based approaches that disregard temporal information (Omarchevska et al., [<reflink idref="bib50" id="ref76">50</reflink>]). ENA has been extensively employed in learning sciences, enabling comparisons of co-occurrence networks among diverse groups or stages (e.g., Ouyang et al., [<reflink idref="bib51" id="ref77">51</reflink>]; Zhang et al., [<reflink idref="bib78" id="ref78">78</reflink>]). For example, Zhang et al. ([<reflink idref="bib78" id="ref79">78</reflink>]) applied ENA to investigate regulatory patterns of student teachers' online collaborative learning activities and compare patterns between high- and low-performing groups across distinct stages of learning activities.</p> <p>In contrast to ENA, which models connections between discursive elements, sequential pattern mining techniques emphasize discovering frequent and significant directed patterns or subsequences within sequential events, behaviors, or educational dialogues (Fournier-Viger et al., [<reflink idref="bib16" id="ref80">16</reflink>]; Song et al., [<reflink idref="bib61" id="ref81">61</reflink>]). Identified ordered patterns reveal the structure or relationship of events or behaviors, helping to understand key characteristics and adjust decision-making strategies accordingly (Vista et al., [<reflink idref="bib71" id="ref82">71</reflink>]). Despite numerous algorithms designed for discovering sequential patterns in sequence databases, lag sequential analysis (LSA) remains one of the most popular sequence mining techniques in education (Song et al., [<reflink idref="bib61" id="ref83">61</reflink>]; Tlili et al., [<reflink idref="bib65" id="ref84">65</reflink>]). Specifically, LSA calculates conditional probabilities of different events and determines the likelihood of an event occurring at a specific time lag after another event (Bakeman & Gottman, [<reflink idref="bib1" id="ref85">1</reflink>]), providing insights into their causal relationships. LSA has been widely used to identify statistically significant patterns in classroom dialogue (e.g., Ma et al., [<reflink idref="bib40" id="ref86">40</reflink>]), online discussion (e.g., Wu et al., [<reflink idref="bib75" id="ref87">75</reflink>]), and collaborative argumentation (e.g., Gao et al., [<reflink idref="bib18" id="ref88">18</reflink>]).</p> <hd id="AN0189590664-7">Method</hd> <p></p> <hd id="AN0189590664-8">Data Collection</hd> <p>Data were collected from an online one-on-one tutoring platform called "Snapask," where university students and individuals with graduate diplomas served as amateur tutors. These individuals enter the field with more knowledge and skills than the students they tutor, but lack certified teaching licenses or significant prior teaching experience. Tutees (i.e., K-12 students) requiring assistance would submit pictures of their questions and provide basic information, such as their educational levels and subjects. The platform automatically matched students with suitable tutors, who then engaged in tutoring sessions to discuss the questions. To account for the potential influence of different payment methods (i.e., free, token-based, and regular subscription) on tutor–student interactions, we exclusively selected dialogues from students who paid through regular subscriptions. Given the substantial number of student questions related to mathematics and the unique context where a mathematics tutor might address queries from primary, middle, and high school students, our study focused on collecting and analyzing tutoring dialogues from these mathematics tutors.</p> <p>Specifically, "Snapask" provided us with a dataset of tutoring records in Singapore from their platform. The dataset included tutor ID, student ID, students' grade levels, dialogue ID, the content of each message, and other features. To identify tutors who interacted with primary school students as well as middle and high school students, we filtered the dataset based on the grade levels of the students with whom they engaged. As a result, we identified 56 such tutors and selected 500 tutoring dialogues from each educational level, namely primary, middle, and high school, resulting in a total of 1500 dialogues (i.e., 22,853 messages). Of the 56 tutors, 15 identified as male and 8 as female, while the remaining tutors did not disclose their gender. Among the 176 students involved in the dialogues, 63 were from primary schools, 63 from middle schools, and the remaining 50 students were from high schools. Table 1 presents a concise overview of the selected tutoring dialogues. The number of tutor messages in each dialogue consistently exceeds that of student messages at all three educational levels. Additionally, both students and tutors exhibit the highest number of utterances at the high school level.</p> <p>Table 1 Description of the dialogues between tutors and students from the primary, middle, and high school levels</p> <p> <ephtml> <table rules="groups"><thead><tr><th align="left" /><th align="left" colspan="2"><p>Message in each dialogue</p></th><th align="left" colspan="2"><p>Student messages in each dialogue</p></th><th align="left" colspan="2"><p>Tutor message in each dialogue</p></th></tr><tr><th align="left" /><th align="left"><p>Mean</p></th><th align="left"><p>SD</p></th><th align="left"><p>Mean</p></th><th align="left"><p>SD</p></th><th align="left"><p>Mean</p></th><th align="left"><p>SD</p></th></tr></thead><tbody><tr><td align="left"><p>Primary</p></td><td char="." align="char"><p>11.492</p></td><td char="." align="char"><p>15.994</p></td><td char="." align="char"><p>3.750</p></td><td char="." align="char"><p>6.592</p></td><td char="." align="char"><p>7.742</p></td><td char="." align="char"><p>9.912</p></td></tr><tr><td align="left"><p>Middle</p></td><td char="." align="char"><p>13.102</p></td><td char="." align="char"><p>17.158</p></td><td char="." align="char"><p>3.924</p></td><td char="." align="char"><p>6.879</p></td><td char="." align="char"><p>9.178</p></td><td char="." align="char"><p>11.153</p></td></tr><tr><td align="left"><p>High</p></td><td char="." align="char"><p>21.110</p></td><td char="." align="char"><p>34.671</p></td><td char="." align="char"><p>8.156</p></td><td char="." align="char"><p>15.023</p></td><td char="." align="char"><p>12.954</p></td><td char="." align="char"><p>20.499</p></td></tr></tbody></table> </ephtml> </p> <hd id="AN0189590664-9">Data Annotation</hd> <p>To analyze how tutors interact with students from different educational levels in the context of online one-on-one mathematics tutoring, we annotated each message within the tutoring dialogues using a coding scheme specifically designed for online one-on-one tutoring and a well-trained artificial intelligence (AI) model.</p> <hd id="AN0189590664-10">Coding Scheme for Online One-on-One Tutoring Dialogue</hd> <p>Primarily drawing upon the ICAP framework (Chi & Wylie, [<reflink idref="bib9" id="ref89">9</reflink>]) and the work of Vail and Boyer ([<reflink idref="bib70" id="ref90">70</reflink>]), we have developed a coding scheme for analyzing online one-on-one tutoring dialogues, which is presented in Table 2.</p> <p>Table 2 The coding scheme for online one-on-one tutoring dialogue</p> <p> <ephtml> <table rules="groups"><thead><tr><th align="left"><p>Group</p></th><th align="left"><p>Name</p></th><th align="left"><p>Code</p></th><th align="left"><p>Role</p></th><th align="left"><p>Description</p></th></tr></thead><tbody><tr><td align="left" rowspan="3"><p>Question asking</p></td><td align="left"><p>Constructive question</p></td><td align="left"><p>CQ</p></td><td align="left"><p>T&S</p></td><td align="left"><p>Ask for ideas and opinions that can lead to task resolution</p></td></tr><tr><td align="left"><p>Procedural question</p></td><td align="left"><p>PQ</p></td><td align="left"><p>T&S</p></td><td align="left"><p>Ask specifically about the procedure, process, or a series of steps for solving a task or achieving a certain outcome</p></td></tr><tr><td align="left"><p>Reasoning question</p></td><td align="left"><p>REQ</p></td><td align="left"><p>T&S</p></td><td align="left"><p>Ask for logical reasoning to prove, justify, or infer a cause-and-effect relationship with evidence or examples</p></td></tr><tr><td align="left" rowspan="3"><p>Evaluation and feedback</p></td><td align="left"><p>Evaluation inducing question</p></td><td align="left"><p>EVQ</p></td><td align="left"><p>T&S</p></td><td align="left"><p>Ask the listener to evaluate the speaker's expressions</p></td></tr><tr><td align="left"><p>Agreement</p></td><td align="left"><p>A</p></td><td align="left"><p>T&S</p></td><td align="left"><p>Affirm agreement or confirm correctness</p></td></tr><tr><td align="left"><p>Neutral and negative feedback</p></td><td align="left"><p>FB</p></td><td align="left"><p>T&S</p></td><td align="left"><p>Provide neutral or negative feedback to the listener's expression</p></td></tr><tr><td align="left" rowspan="4"><p>Constructive participation</p></td><td align="left"><p>Constructive expression</p></td><td align="left"><p>C</p></td><td align="left"><p>S</p></td><td align="left"><p>Provide ideas or opinions that advance the resolution of the task</p></td></tr><tr><td align="left"><p>Solution procedure</p></td><td align="left"><p>P</p></td><td align="left"><p>S</p></td><td align="left"><p>Provide the procedure or process of the whole or partial solution, achieving certain outcomes</p></td></tr><tr><td align="left"><p>Reasoning</p></td><td align="left"><p>RE</p></td><td align="left"><p>S</p></td><td align="left"><p>Explain logical reasoning, justifications, or explanations with evidence or examples</p></td></tr><tr><td align="left"><p>Knowledge sharing and instruction</p></td><td align="left"><p>KS</p></td><td align="left"><p>T</p></td><td align="left"><p>Provide information or instruction, share knowledge or explain solutions</p></td></tr><tr><td align="left"><p>Active participation</p></td><td align="left"><p>Acknowledgment</p></td><td align="left"><p>AP</p></td><td align="left"><p>S</p></td><td align="left"><p>Simple acknowledgment without additional constructive action</p></td></tr><tr><td align="left" rowspan="3"><p>Metacognition</p></td><td align="left"><p>Reflection inducing question</p></td><td align="left"><p>RFQ</p></td><td align="left"><p>T&S</p></td><td align="left"><p>Students are asked to reflect on their understanding, performance, or progress (mainly by tutors)</p></td></tr><tr><td align="left"><p>Other reflection</p></td><td align="left"><p>RF</p></td><td align="left"><p>T&S</p></td><td align="left"><p>Reflection on understanding, performance, or progress (mainly by students themselves)</p></td></tr><tr><td align="left"><p>Reflection of unknowing</p></td><td align="left"><p>RFN</p></td><td align="left"><p>T&S</p></td><td align="left"><p>Reflecting a partial or a lack of understanding</p></td></tr><tr><td align="left"><p>Off-topic</p></td><td align="left"><p>Off-topic chatting</p></td><td align="left"><p>OT</p></td><td align="left"><p>T&S</p></td><td align="left"><p>Off-topic expressions or questions</p></td></tr></tbody></table> </ephtml> </p> <p> <emph>Note. </emph>In the formal analysis, we use S.xx and T.xx to represent the code for teacher and student messages</p> <p>The coding scheme comprises two layers. The first layer consists of six categories: question asking, evaluation and feedback, constructive participation, active participation, metacognition, and off-task behavior. The categories of question asking and evaluation and feedback align with the interactive behavior described in the ICAP framework, while constructive participation and active participation correspond to the constructive and active dimensions, respectively. Given the nature of online one-on-one tutoring, which often involves metacognitive behaviors such as checking student understanding of solutions, we included a metacognitive dimension in the coding scheme. Furthermore, we introduced a category called "off-topic" to capture instances of frequent off-topic dialogue. The second layer encompasses 15 dialogic acts, each associated with one of the first-layer categories. For example, the category of question asking includes subcategories such as constructive questions (seeking opinions), procedural questions (inquiring about solution procedures), and reasoning questions (seeking explanations).</p> <p>It is worth noting that certain codes can be used to annotate both tutors' and students' messages. For instance, constructive questions, procedural questions, and reasoning questions can be posed by both tutors and students. However, some dialogic acts are exclusive to either tutors or students. For instance, only students can acknowledge others' expressions, which is referred to as "Acknowledgment," while "knowledge sharing and instruction" can only be performed by tutors.</p> <hd id="AN0189590664-11">AI Model for Data Annotation</hd> <p>We employed a well-trained AI model to automatically annotate the large number of online one-on-one tutoring dialogues. Initially, we manually annotated 593 tutoring sessions, comprising over 1500 messages. Subsequently, we trained three AI models, namely Bert, Roberta, and Bertweet, and evaluated their performance to select the most effective model for automatic annotation. Bert (Devlin et al., [<reflink idref="bib13" id="ref91">13</reflink>]) was chosen for its effectiveness in capturing dependencies between utterances in dialogues, as it is trained for next sentence prediction. Roberta (Liu et al., [<reflink idref="bib39" id="ref92">39</reflink>]), a variant of Bert with a different masking strategy and trained on more data, was also considered. Additionally, Bertweet (Nguyen et al., [<reflink idref="bib48" id="ref93">48</reflink>]), a variant of Roberta pre-trained on tweets, was selected for its exposure to spelling errors and informal expressions.</p> <p>To ensure a robust evaluation, we employed a five-fold cross-validation method to train the three models. The Bert model achieved a micro F1 score of 0.886 and a macro F1 score of 0.758, while the Roberta model achieved a micro F1 score of 0.890 and a macro F1 score of 0.767. The Bertweet model achieved a micro F1 score of 0.866 and a macro F1 score of 0.719. Considering the outstanding performance of the Roberta model, we selected it for annotating the collected tutoring dialogues. Furthermore, we conducted a validation study by manually annotating 100 randomly selected sessions and the consistency between manual and automatic annotations achieved an F1 score of 0.87, affirming the effectiveness of the automatic model in annotating the data.</p> <hd id="AN0189590664-12">Data Analysis</hd> <p>Firstly, to examine the most and least utilized dialogic strategies by tutors and students at different educational levels and explore the patterns of co-occurrence among these strategies, we utilized the ENA approach. Specifically, we utilized the ENA Web Tool, accessible at https://app.epistemicnetwork.org, which facilitated the analysis and visualization of dialogic strategy occurrences and co-occurrences through a user-friendly graph-based network representation. Prior to conducting ENA, we established four key components: codes, conversations, the unit of analysis, and the stanza. Codes represented the specific dialogic strategies (i.e., dialogic acts in our coding scheme) that each message was annotated with. Conversations indicated the turns of talk exchanged within a specific activity, referring to the utterances between students and tutors during a tutoring session in our study. The unit of analysis was further categorized based on students' educational levels, including primary, middle, and high school. Lastly, the stanza referred to the range of co-occurrence between codes modeled by ENA, with a moving stanza window of 4 set for this study. In the visualized ENA networks, codes were represented as nodes, with their size indicating the prevalence. Larger nodes corresponded to more frequent occurrences of the associated code, enabling us to understand the frequent and infrequent dialogic strategies in online one-on-one tutoring. The connections between codes were depicted as edges, with the thickness of an edge representing the frequency of co-occurrence between the linked codes. Thicker edges indicated a higher rate of co-occurrence, indicating a stronger association between the respective dialogic strategies and enabling us to understand how tutors leveraged these strategies to facilitate tutoring.</p> <p>Secondly, in addition to ENA, which captures occurrences and co-occurrences of dialogic strategies, we also employed LSA to uncover statistically significant sequential patterns of how tutors interact with students across primary, middle, and high school levels. While ENA provides an intuitive description of the co-occurrences, it does not assess the statistical significance of these associations. LSA, on the other hand, allows us to determine the significance of one dialogic strategy following another, thus revealing causal relationships. By incorporating LSA as a complementary method to ENA, we conducted a comprehensive analysis of how amateur mathematics tutors engage with students across different educational levels. Similar to ENA, the results of LSA were visualized in a graph format consisting of nodes and arrows (representing directed edges). Nodes in the LSA graph also represent dialogic strategies. An arrow originating from one node (e.g., strategy A) and pointing to another (e.g., strategy B) indicates the probability of strategy B following strategy A is statistically significant, suggesting utilizing strategy A may cause the subsequent use of strategy B. The values close to the arrows and the arrow thickness respectively represented the z-score and significance level of the sequential relationship.</p> <hd id="AN0189590664-13">Results</hd> <p></p> <hd id="AN0189590664-14">Epistemic Network Analysis</hd> <p></p> <hd id="AN0189590664-15">Occurrences of Individual Dialogic Strategies</hd> <p>Figures 1, 2, and 3 illustrate the overall networks of dialogic strategies among tutors' interactions with primary, middle, and high school students in online one-on-one tutoring, where the node size indicates the frequency of these strategies. The red, blue, and purple dots represent the centroids of tutors and primary, middle, and high school students, respectively. Moreover, Fig. 4 displays the distribution of tutors' dialogic strategies in this context.</p> <p>Graph: Fig. 1 The overall networks of dialogic interactions between tutors and students at the primary school level</p> <p>Graph: Fig. 2 The overall networks of dialogic interactions between tutors and students at the middle school level</p> <p>Graph: Fig. 3 The overall networks of dialogic interactions between tutors and students at the high school level</p> <p>Graph: Fig. 4 The distribution of tutors' dialogic strategies when they interact with students from primary, middle, and high schools in online one-on-one tutoring</p> <p>More specifically, the node sizes of T.OT (off-topic), T.KS (knowledge sharing and instruction), and T.RFQ (reflection inducing question) are consistently the most prominent dialogic strategies in all three networks (primary, middle, and high). This finding suggests that tutors often engage in off-task dialogues, which form a considerable part of their interactions. Notably, Fig. 4 reveals that tutors used a higher proportion of off-topic utterances when interacting with primary and middle school students than with high school students. Tutors mainly employed knowledge sharing and instruction strategies (T.KS), particularly when tutoring high school students. Encouraging student reflection (T.RFQ) was another common behavior observed in their tutoring. Additionally, tutors utilized T.CQ (constructive questions), T.A (agreement), and T.FB (neutral and negative feedback) when instructing students across different grade levels, albeit at lower frequencies. Furthermore, students of all educational levels exhibited a substantial amount of off-topic behavior (S.OT) during online tutoring sessions.</p> <hd id="AN0189590664-16">Co-occurrences of Different Dialogic Strategies</hd> <p>The thickness of lines between the two codes in Figs. 1, 2, and 3 represents the co-occurrence of tutors' use of different dialogic strategies when interacting with students from various educational levels. Five predominant patterns are observed across all grade levels: T.OT and T.KS, S.OT and T.OT, S.OT and T.KS, T.KS and T.RFQ, and T.OT and T.RFQ. Excluding the co-occurrence of tutors' off-topic behavior (i.e., T.OT) with other dialogic strategies, T.KS (knowledge sharing and instruction) and T.RFQ (reflection inducing question) are the only dominant pattern shared by the three educational levels. Additionally, more co-occurrences of dialogic strategies seem to appear in tutors' interactions with high school students, such as T.KS and S.C.</p> <p>To further illustrate the differences in tutors' dialogic strategies when tutoring primary, middle, and high school students, Figs. 5 and 6 present the subtracted networks derived from the overall networks. These networks display distinct network graphs by subtracting nodes and connection weights from two overall networks (i.e., comparing the middle and primary school networks and comparing the middle and high school networks). The dominance of the blue color in the subtracted networks indicates that tutors and middle school students had a greater number of co-occurrences of the specified codes compared to tutors and primary school or high school students, and vice versa. These subtracted networks provide a visual representation of the specific features and patterns that distinguish the dialogic interactions within each educational level (Eagan & Hamilton, [<reflink idref="bib14" id="ref94">14</reflink>]).</p> <p>Graph: Fig. 5 The subtracted networks showing differences in the patterns of dialogic strategies between tutorship for middle school students and for primary school students</p> <p>Graph: Fig. 6 The subtracted networks showing differences in the patterns of dialogic strategies between tutorship for middle school students and for high school students</p> <p>After filtering out the off-topic codes, Table 3 presents the coefficients of the main connections in the subtracted networks, enabling a comparison of the strength of co-occurrence between different dialogic strategies during tutorship across various educational levels. Notably, the correlations between tutors' constructive questions (T.CQ) and students' constructive expression (S.C) differ based on the students' educational level. Specifically, the link between T.CQ and S.C is stronger in tutorship for high school students than in tutorship for middle school students. However, no significant difference in this correlation is observed between tutorship for primary school students and tutorship for middle school students. These results suggest that when tutoring higher-grade-level students, tutors tend to employ constructive questions more frequently, and these questions may be more likely to elicit greater constructive expressions from high school students.</p> <p>Table 3 Connection coefficients for the subtracted networks of tutors and primary/middle/high students</p> <p> <ephtml> <table rules="groups"><thead><tr><th align="left"><p>Connection</p></th><th align="left"><p>Subtracted weight (middle − primary)</p></th><th align="left"><p>Subtracted weight (middle − high)</p></th></tr></thead><tbody><tr><td align="left"><p>T.CQ-S.C</p></td><td align="left"><p>/</p></td><td align="left"><p> − 0.01(middle < high)</p></td></tr><tr><td align="left"><p>T.KS-S.C</p></td><td align="left"><p>0.02(middle > primary)</p></td><td align="left"><p> − 0.1(middle < high)</p></td></tr><tr><td align="left"><p>T.KS-S.AP</p></td><td align="left"><p> − 0.01(middle < primary)</p></td><td align="left"><p>/</p></td></tr></tbody></table> </ephtml> </p> <p> <emph>T.KS</emph> tutors' knowledge sharing and instruction, <emph>T.CQ</emph> tutors' constructive question, <emph>S.AP</emph> students' acknowledgment, <emph>S.C</emph> students' constructive expression</p> <p>Furthermore, the correlations between tutors' knowledge sharing and instruction (T.KS) and students' constructive expression (S.C) during tutorship differ across educational levels. Specifically, these correlations are stronger in tutorship for high school students than in tutorship for middle school students. Moreover, the correlations in tutorship for middle school students are stronger than those in tutorship for primary school students. In contrast, the correlations between tutors' knowledge sharing and instruction (T.KS) and students' acknowledgment (S.AP) in tutorship for primary school students are stronger than those in tutorship for middle school students. These findings suggest that tutors' knowledge sharing and instruction strategies may promote more advanced constructive expressions among students at higher educational levels. However, for primary school students, these strategies may primarily result in simple acknowledgment without further constructive action.</p> <hd id="AN0189590664-17">Lag Sequential Analysis</hd> <p>Figures 7, 8, and 9 illustrate the statistically significant sequential patterns of tutors' dialogic strategies during one-on-one online tutoring sessions with students from different educational levels. To highlight the differences, we used black arrows to represent common patterns shared by all three education levels, while red, blue, and purple arrows indicate unique dialogic strategy sequences for primary, middle, and high school students, respectively.</p> <p>Graph: Fig. 7 Transition diagram of tutors' dialogic strategies when tutoring primary school students in online one-on-one tutoring</p> <p>Graph: Fig. 8 Transition diagram of tutors' dialogic strategies when tutoring middle school students in online one-on-one tutoring</p> <p>Graph: Fig. 9 Transition diagram of tutors' dialogic strategies when tutoring high school students in online one-on-one tutoring</p> <p>Tutors' interactions with students from the three educational levels reveal four common patterns: T.OT → T.OT, T.KS → T.KS, T.KS → T.RFQ, and T.FB → T.REQ. The first three patterns align with the results of ENA, indicating tutors' frequent off-topic behavior, as well as the consecutive use of knowledge sharing and instruction (T.KS) strategies and reflection-inducing questions (T.RFQ). Furthermore, tutors tend to pose reasoning questions (T.REQ) following neutral and negative feedback (FB) across all three educational levels during online tutoring.</p> <p>Tutors also demonstrate unique patterns of dialogic strategies when interacting with students from different grades. For primary and middle school students, tutors exhibit relatively limited dialogic strategy patterns, predominantly involving question-asking behaviors and evaluation and feedback. For instance, during primary student tutoring, tutors often ask constructive or procedural questions, followed by evaluation-inducing questions (T.CQ/T.PQ → T.EVQ). In middle school tutoring, tutors frequently repeat constructive questions (T.CQ → T.CQ), reasoning questions (T.REQ → T.REQ), and evaluation-inducing questions (T.EVQ → T.EVQ). However, they seem to lack the use of metacognition-related strategies.</p> <p>In contrast, tutors display more complex dialogic strategy patterns when tutoring high school students, which is consistent with the ENA results. Tutors effectively employ strategies across various dimensions, including question-asking, evaluation and feedback, constructive participation, and metacognition. Notably, almost half of the patterns involve T.REQ (reasoning questions), indicating a strong emphasis on reasoning for high school students. Additionally, tutors concentrate on fostering high school students' metacognition during tutoring sessions, as demonstrated by the frequent occurrence of metacognition-related patterns. For example, tutors tend to request students to reflect on their knowing or unknowing after asking reasoning questions (T.REQ → T.RF/T.RFN).</p> <hd id="AN0189590664-18">Discussion and Conclusion</hd> <p></p> <hd id="AN0189590664-19">Summary of Findings</hd> <p>This study investigates the interaction dynamics between tutors and students at different educational levels during online one-on-one mathematics tutoring. We collected 1500 tutoring dialogues from amateur mathematics tutors who were simultaneously instructing students from primary, middle, and high school levels. These dialogues were annotated using a coding scheme specifically designed for online one-on-one tutoring and a well-trained powerful AI model, with each message being categorized. The interaction dynamics were then examined using ENA, which allowed us to analyze the occurrences and co-occurrences of dialogic strategies. Additionally, LSA was employed to identify statistically significant patterns among the co-occurrences of these strategies. Insightful findings were obtained.</p> <p>Firstly, the statistical analysis reveals a common thread across all educational levels that tutors always dominate the tutoring sessions, with their messages accounting for the majority of the conversation (67.4% for primary school level, 70.1% for middle school level, and 61.4% for high school level). This finding is consistent with previous research on offline one-on-one tutoring (Chi et al., [<reflink idref="bib8" id="ref95">8</reflink>]; Graesser et al., [<reflink idref="bib21" id="ref96">21</reflink>]) and teacher-dominated mathematics lessons (Mameli et al., [<reflink idref="bib41" id="ref97">41</reflink>]). However, a closer examination reveals a difference in the number of messages in a tutoring dialogue, which increases with the educational level. Specifically, primary and middle school students contribute significantly fewer messages compared to tutors, which may suggest a more passive learning approach at these stages. In contrast, high school students are more actively engaged, as evidenced by their higher number of messages, suggesting a shift towards a more interactive and meaningful dialogue.</p> <p>Secondly, the results of ENA reveal consistent behaviors among tutors across all educational levels, with off-topic messages (T.OT), knowledge sharing and instruction (T.KS), and reflection inducing questions (T.RFQ) being the most frequent behaviors. The prevalence of off-topic messages suggests that these amateur tutors may lack teaching experience and training, leading to inefficient instruction, as pointed out by previous research (e.g., Booth et al., [<reflink idref="bib4" id="ref98">4</reflink>]; Nickow et al., [<reflink idref="bib49" id="ref99">49</reflink>]; Reichenberg et al., [<reflink idref="bib52" id="ref100">52</reflink>]). However, some researchers found that casual interactions can also contribute to building rapport and enhancing students' comfort (e.g., Tu & McIsaac, [<reflink idref="bib68" id="ref101">68</reflink>]). The majority of knowledge sharing and instruction (T.KS) behavior in tutors' instruction indicates that tutors prefer direct instruction over other dialogic strategies, aligning with previous findings (Hrastinski et al., [<reflink idref="bib30" id="ref102">30</reflink>]). However, successful tutoring relies on scaffolding (Chi et al., [<reflink idref="bib8" id="ref103">8</reflink>]); thus, tutors should focus on facilitating discourse rather than providing direct answers, as recommended by Hrastinski et al. ([<reflink idref="bib30" id="ref104">30</reflink>]) and Garrison et al. ([<reflink idref="bib19" id="ref105">19</reflink>]). Additionally, tutors also pose reflection inducing questions (T.RFQ) to promote students' reflection, albeit to a lesser extent.</p> <p>Thirdly, the results of ENA illuminate the co-occurring dialogic strategies during online one-on-one tutoring, revealing both commonalities and distinctions across educational levels. Notably, regardless of the educational level, tutors tend to concurrently utilize five fixed combinations of dialogic strategies centering around T.OT, T.KS, and T.RFQ (i.e., T.OT and T.KS, T.OT and S.OT, T.OT and T.RFQ, T.KS and S.OT, and T.KS and T.RFQ). This potentially indicates tutors' lack of proficiency in applying different effective dialogic skills, as pointed out by previous research (Correnti et al., [<reflink idref="bib11" id="ref106">11</reflink>]; Feng et al., [<reflink idref="bib15" id="ref107">15</reflink>]; McFarlane, [<reflink idref="bib42" id="ref108">42</reflink>]; Trygstad et al., [<reflink idref="bib66" id="ref109">66</reflink>]). Filtering out off-topic patterns, the most common co-occurrence involves knowledge sharing and instruction (T.KS) and reflection inducing questions (T.RFQ), highlighting the main similarity in tutors' instruction across different educational levels. However, the analysis also uncovers the differences in how students at different levels engage with these strategies. Specifically, the analysis shows that high school students exhibit more advanced constructive expressions (S.C) associated with tutors' knowledge sharing and instruction (T.KS) and constructive questions (T.CQ), while primary students primarily show simple acknowledgment (S.AP). This finding highlights the potential challenges faced by primary school students in articulating complex ideas or opinions that contribute to task resolution. As a result, it may be advantageous for tutors to adopt tailored approaches that encourage primary school students to move beyond mere acknowledgment and actively participate in constructive discussions.</p> <p>Fourthly, the results of LSA identify statistically significant sequential patterns within the co-occurring dialogic strategies across tutors' interactions with students at different educational levels. In addition to the patterns discovered by ENA (e.g., T.KS → T.RFQ), it is found that tutors consistently tend to pose reasoning questions (T.REQ) following neutral and negative feedback (FB) across all educational levels, indicating their efforts to highlight reasoning. However, the analysis also highlights a difference in the variety of dialogic strategies employed by tutors when interacting with students of different educational levels. When tutoring primary and middle school students, tutors exhibit limited dialogic strategy patterns, mainly focusing on question asking behaviors and evaluation and feedback behaviors. In contrast, when instructing high school students, tutors show diverse patterns across all dimensions of the coding scheme, with a specific emphasis on facilitating students' reasoning and metacognition. These findings suggest that while tutors attempt to adopt different dialogic strategies for high school students, their interaction with primary and middle school students requires further improvement and more tailored training, such as focusing on students' metacognition, as recommended by Graesser et al. ([<reflink idref="bib22" id="ref110">22</reflink>]) and Hrastinski et al. ([<reflink idref="bib30" id="ref111">30</reflink>]).</p> <p>In summary, the analysis of tutoring dialogues between tutors and students reveals both similarities and differences across different educational levels. A common thread is the tutors' dominant role and the display of similar behavior patterns (i.e., T.OT, T.KS, T.RFQ, T.REQ, and FB) in tutoring at all three educational levels, indicating a need for tutors to enhance their diverse dialogic skills. Compared with high school students' advanced constructive expressions in response to tutors' instruction, primary school students seem to fail to engage in constructive discussions spontaneously. Therefore, tutors are supposed to develop more advanced dialogic strategies to promote primary school students' meaningful learning. Additionally, tutors display a greater diversity of dialogic strategies for high school students, often omitting the facilitation of metacognition for primary and middle school students. These findings provide a nuanced picture of online one-on-one tutoring dialogues, informing the development of tutors' dialogic teaching skills.</p> <hd id="AN0189590664-20">Limitations and Future Research Directions</hd> <p>Although the study offers valuable insights, several limitations must be acknowledged. Firstly, the data were derived from amateur mathematics tutors who instructed students at varying educational levels in Singapore. Consequently, the generalizability of these findings to other contexts, such as different subjects, regions, and types of tutors, remains uncertain. Secondly, our dataset comprised 1500 text-based tutoring dialogues (500 from each educational level), totaling 22,853 messages. A more extensive and multimodal dataset could potentially yield a more comprehensive understanding and analysis of interaction dynamics in online one-on-one tutoring. Thirdly, the employed coding scheme may not encompass all dialogic behaviors present in online one-on-one tutoring and may require further refinement or enhancement.</p> <p>Future research should, therefore, examine online one-on-one tutoring within a broader context, utilizing more exhaustive and multimodal data, as well as an improved coding scheme. Furthermore, building upon the findings of this study, future research could also develop and implement training programs aimed at enhancing tutors' professional capabilities during instruction.</p> <hd id="AN0189590664-21">Author Contribution</hd> <p>Deliang Wang and Lei Gao: conceptualization, methodology, investigation, formal analysis, writing—original draft, writing—reviewing and editing; Dapeng Shan, Gaowei Chen, Chenwei Zhang, and Ben Kao: methodology, reviewing and editing.</p> <hd id="AN0189590664-22">Data Availability</hd> <p>The data and materials that support the findings of this study are available from the first author, Deliang Wang, upon reasonable request.</p> <hd id="AN0189590664-23">Declarations</hd> <p></p> <hd id="AN0189590664-24">Ethical Approval</hd> <p>Ethical approval was issued by the Human Research Ethics Committee of the University of Hong Kong.</p> <hd id="AN0189590664-25">Consent to Participate</hd> <p>Not applicable.</p> <hd id="AN0189590664-26">Consent for Publication</hd> <p>Not applicable.</p> <hd id="AN0189590664-27">Competing Interests</hd> <p>The authors declare no competing interests.</p> <hd id="AN0189590664-28">Publisher's Note</hd> <p>Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.</p> <ref id="AN0189590664-29"> <title> References </title> <blist> <bibl id="bib1" idref="ref85" type="bt">1</bibl> <bibtext> Bakeman R, Gottman JM. Observing interaction: An introduction to sequential analysis. 1997; Cambridge University Press. 10.1017/CBO9780511527685</bibtext> </blist> <blist> <bibl id="bib2" idref="ref23" type="bt">2</bibl> <bibtext> Bates SP, Galloway RK, Riise J, Homer D. Assessing the quality of a student-generated question repository. Physical Review Special Topics-Physics Education Research. 2014; 10; 2: 020105. 10.1103/PhysRevSTPER.10.020105</bibtext> </blist> <blist> <bibl id="bib3" idref="ref4" type="bt">3</bibl> <bibtext> Bloom BS. The 2 sigma problem: The search for methods of group instruction as effective as one-to-one tutoring. Educational Researcher. 1984; 13; 6: 4-16. 10.3102/0013189X013006004</bibtext> </blist> <blist> <bibl id="bib4" idref="ref25" type="bt">4</bibl> <bibtext> Booth, B. M, Jacobs, J, Bush, J. B, Milne, B, Fischaber, T, & D'Mello, S. K. (2024). Human-tutor coaching technology (htct): Automated discourse analytics in a coached tutoring model. The 14th Learning Analytics and Knowledge Conference (LAK'24), March 18–22, 2024, Kyoto, Japan.</bibtext> </blist> <blist> <bibl id="bib5" idref="ref62" type="bt">5</bibl> <bibtext> Boyer, K, Phillips, R, Wallis, M, Vouk, M, & Lester, J. (2008). Learner characteristics and feedback in tutorial dialogue. Proceedings of the Third Workshop on Innovative Use of NLP for Building Educational Applications (pp. 53–61).</bibtext> </blist> <blist> <bibl id="bib6" idref="ref17" type="bt">6</bibl> <bibtext> Chappell S, Arnold P, Nunnery J, Grant M. An examination of an online tutoring program's impact on low-achieving middle school students' mathematics achievement. Online Learning. 2015; 19; 5: 37-53. 10.24059/olj.v19i5.694</bibtext> </blist> <blist> <bibl id="bib7" idref="ref41" type="bt">7</bibl> <bibtext> Chen G. A visual learning analytics (vla) approach to video-based teacher professional development: Impact on teachers' beliefs, self-efficacy, and classroom talk practice. Computers & Education. 2020; 144. 10.1016/j.compedu.2019.103670103670</bibtext> </blist> <blist> <bibl id="bib8" idref="ref35" type="bt">8</bibl> <bibtext> Chi MT, Siler SA, Jeong H, Yamauchi T, Hausmann RG. Learning from human tutoring. Cognitive Science. 2001; 25; 4: 471-533. 10.1207/s15516709cog2504_1</bibtext> </blist> <blist> <bibl id="bib9" idref="ref51" type="bt">9</bibl> <bibtext> Chi MT, Wylie R. The icap framework: Linking cognitive engagement to active learning outcomes. Educational Psychologist. 2014; 49; 4: 219-243. 10.1080/00461520.2014.965823</bibtext> </blist> <blist> <bibtext> Chin C. Teacher questioning in science classrooms: Approaches that stimulate productive thinking. Journal of Research in Science Teaching: The Official Journal of the National Association for Research in Science Teaching. 2007; 44; 6: 815-843. 10.1002/tea.20171</bibtext> </blist> <blist> <bibtext> Correnti R, Stein MK, Smith MS, Scherrer J, McKeown M, Greeno J, Ashley K. Improving teaching at scale: Design for the scientific measurement and learning of discourse practice. 2015: American Educational Research Association; Socializing intelligence through academic talk and dialogue: 303-320</bibtext> </blist> <blist> <bibtext> De Ree J, Maggioni MA, Paulle B, Rossignoli D, Ruijs N, Walentek D. Closing the income-achievement gap? Experimental evidence from high-dosage tutoring in Dutch primary education. Economics of Education Review. 2023; 94: 102383. 10.1016/j.econedurev.2023.102383</bibtext> </blist> <blist> <bibtext> Devlin, J, Chang, M.-W, Lee, K, & Toutanova, K. (2018). Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805.</bibtext> </blist> <blist> <bibtext> Eagan, B, & Hamilton, E. (2018). Epistemic network analysis of an international digital makerspace in Africa, Europe, and the US. In Annual Meeting of the American Education Research Association, New York City.</bibtext> </blist> <blist> <bibtext> Feng X, Xie J, Liu Y. Using the community of inquiry framework to scaffold online tutoring. International Review of Research in Open and Distributed Learning. 2017; 18; 2: 162-188. 10.19173/irrodl.v18i2.2362</bibtext> </blist> <blist> <bibtext> Fournier-Viger P, Lin JCW, Kiran RU, Koh YS, Thomas R. A survey of sequential pattern mining. Data Science and Pattern Recognition. 2017; 1; 1: 54-77</bibtext> </blist> <blist> <bibtext> Fuchs LS, Seethaler PM, Powell SR, Fuchs D, Hamlett CL, Fletcher JM. Effects of preventative tutoring on the mathematical problem solving of third-grade students with math and reading difficulties. Exceptional Children. 2008; 74; 2: 155-173. 10.1177/001440290807400202</bibtext> </blist> <blist> <bibtext> Gao L, Li X, Li Y, Hu W. Capturing temporal and sequential patterns of socio-emotional interaction in high-and low-performing collaborative argumentation groups. The Asia-Pacific Education Researcher. 2023; 32; 6: 817-831. 10.1007/s40299-022-00698-7</bibtext> </blist> <blist> <bibtext> Garrison DR, Anderson T, Archer W. Critical inquiry in a text-based environment: Computer conferencing in higher education. The Internet and Higher Education. 1999; 2; 2–3: 87-105. 10.1016/S1096-7516(00)00016-6</bibtext> </blist> <blist> <bibtext> Graesser AC, D'Mello S, Cade WMayer RE, Alexander PA. Handbook of research on learning and instruction. 2011; Routledge Press: 408-426</bibtext> </blist> <blist> <bibtext> Graesser AC, Person NK, Magliano JP. Collaborative dialog patterns in naturalistic one-to-one tutoring. Applied Cognitive Psychology. 1995; 9; 6: 495-522. 10.1002/acp.2350090604</bibtext> </blist> <blist> <bibtext> Graesser, A. C, VanLehn, K, Rosé, C. P, Jordan, P. W, & Harter, D. (2001). Intelligent tutoring systems with conversational dialogue. AI magazine, 22(4), 39–39. https://doi.org/10.1609/aimag.v22i4.1591</bibtext> </blist> <blist> <bibtext> Grand View Research. (2024). Online tutoring services market size & trends report, 2030. Retrieved February 1, 2024, from https://<ulink href="http://www.grandviewresearch.com/industry-analysis/online-tutoring-services-market">www.grandviewresearch.com/industry-analysis/online-tutoring-services-market</ulink></bibtext> </blist> <blist> <bibtext> Hampel R, Stickler U. New skills for new classrooms: Training tutors to teach languages online. Computer Assisted Language Learning. 2005; 18; 4: 311-326. 10.1080/09588220500335455</bibtext> </blist> <blist> <bibtext> Hanham J, Lee CB, Teo T. The influence of technology acceptance, academic self-efficacy, and gender on academic achievement through online tutoring. Computers & Education. 2021; 172: 104252. 10.1016/j.compedu.2021.104252</bibtext> </blist> <blist> <bibtext> Hardt D, Nagler M, Rincke J. Tutoring in (online) higher education: Experimental evidence. Economics of Education Review. 2023; 92: 102350. 10.1016/j.econedurev.2022.102350</bibtext> </blist> <blist> <bibtext> Hennessy S, Howe C, Mercer N, Vrikki M. Coding classroom dialogue: Methodological considerations for researchers. Learning, Culture and Social Interaction. 2020; 25: 100404. 10.1016/j.lcsi.2020.100404</bibtext> </blist> <blist> <bibtext> Hennessy S, Rojas-Drummond S, Higham R, Márquez AM, Maine F, Ríos RM, García-Carrión R, Torreblanca O, Barrera MJ. Developing a coding scheme for analysing classroom dialogue across educational contexts. Learning, Culture and Social Interaction. 2016; 9: 16-44. 10.1016/j.lcsi.2015.12.001</bibtext> </blist> <blist> <bibtext> Howe C, Abedin M. Classroom dialogue: A systematic review across four decades of research. Cambridge Journal of Education. 2013; 43; 3: 325-356. 10.1080/0305764X.2013.786024</bibtext> </blist> <blist> <bibtext> Hrastinski S, Cleveland-Innes M, Stenbom S. Tutoring online tutors: Using digital badges to encourage the development of online tutoring skills. British Journal of Educational Technology. 2018; 49; 1: 127-136. 10.1111/bjet.12525</bibtext> </blist> <blist> <bibtext> Hume G, Michael J, Rovick A, Evens M. Hinting as a tactic in one-on-one tutoring. The Journal of the Learning Sciences. 1996; 5; 1: 23-47. 10.1207/s15327809jls0501_2</bibtext> </blist> <blist> <bibtext> Hung M-L. Exploring the perspectives and readiness of undergraduate tutors regarding synchronous online one-to-one tutoring for pupils in rural areas. The Asia Pacific Education Researcher. 2021; 31; 5: 553-561. 10.1007/s40299-021-00607-4</bibtext> </blist> <blist> <bibtext> Joshi, P. (2020). Private schooling and tutoring at scale in South Asia. In P. Sarangapani, & R. Pappu (Eds.), Handbook of Education Systems in South Asia. Global Education Systems. Springer. https://doi.org/10.1007/978-981-13-3309-5_23-1</bibtext> </blist> <blist> <bibtext> Joubert YT, Snyman AM. Challenges experienced with online tutoring in an ODL institution. Progressio. 2017; 39; 1: 126-145. 10.25159/0256-8853/2139</bibtext> </blist> <blist> <bibtext> Kim M-Y, Wilkinson IA. What is dialogic teaching? Constructing, deconstructing, and reconstructing a pedagogy of classroom talk. Learning, Culture and Social Interaction. 2019; 21: 70-86. 10.1016/j.lcsi.2019.02.003</bibtext> </blist> <blist> <bibtext> Lin, J, Rakovic, M, Lang, D, Gasevic, D, & Chen, G. (2022a). Exploring the politeness of instructional strategies from human-human online tutoring dialogues. LAK22: 12th International Learning Analytics and Knowledge Conference, 282–293.</bibtext> </blist> <blist> <bibtext> Lin J, Singh S, Sha L, Tan W, Lang D, Gašević D, Chen G. Is it a good move? Mining effective tutoring strategies from human–human tutorial dialogues. Future Generation Computer Systems. 2022; 127: 194-207. 10.1016/j.future.2021.09.001</bibtext> </blist> <blist> <bibtext> Litman D, Forbes-Riley K. Correlations between dialogue acts and learning in spoken tutoring dialogues. Natural Language Engineering. 2006; 12; 2: 161-176. 10.1017/S1351324906004165</bibtext> </blist> <blist> <bibtext> Liu, Y, Ott, M, Goyal, N, Du, J, Joshi, M, Chen, D, Levy, O, Lewis, M, Zettlemoyer, L, & Stoyanov, V. (2019). Roberta: A robustly optimized bert pretraining approach. arXiv preprint arXiv:1907.11692.</bibtext> </blist> <blist> <bibtext> Ma, X, Xie, Y, Yang, X, Wang, H, Li, Z, & Lu, J. (2024). Teacher-student interaction modes in smart classroom based on lag sequential analysis. Education and Information Technologies (pp. 1–25). https://doi.org/10.1007/s10639-024-12487-4</bibtext> </blist> <blist> <bibtext> Mameli C, Mazzoni E, Molinari L. Patterns of discursive interactions in primary classrooms: An application of social network analysis. Research Papers in Education. 2015; 30; 5: 546-566. 10.1080/02671522.2015.1027727</bibtext> </blist> <blist> <bibtext> McFarlane KJ. Tutoring the tutors: Supporting effective personal tutoring. Active Learning in Higher Education. 2016; 17; 1: 77-88. 10.1177/14697874156167</bibtext> </blist> <blist> <bibtext> McPherson, M, & Nunes, M. B. (2004). The role of tutors as an integral part of online learning support. Retrieved February 2, 2024, from https://eprints.whiterose.ac.uk/999/</bibtext> </blist> <blist> <bibtext> Mercer N. The analysis of classroom talk: Methods and methodologies. British Journal of Educational Psychology. 2010; 80; 1: 1-14. 10.1348/000709909X479853</bibtext> </blist> <blist> <bibtext> Michaels, S, O'Connor, M. C, Hall, M. W, & Resnick, L. B. (2010). Accountable talk® sourcebook. Pittsburg, PA: Institute for Learning University of Pittsburgh. Murphy, PK, Wilkinson, IAG, Soter, AO, Hennessey, MN, & Alexander, JF.</bibtext> </blist> <blist> <bibtext> Michaels S, O'Connor C, Resnick LB. Deliberative discourse idealized and realized: Accountable talk in the classroom and in civic life. Studies in Philosophy and Education. 2008; 27: 283-297. 10.1007/s11217-007-9071-1</bibtext> </blist> <blist> <bibtext> Mitchell, C. M, Ha, E. Y, Boyer, K. E, & Lester, J. C. (2012). Recognizing effective and student-adaptive tutor moves in task-oriented tutorial dialogue. Twenty-Fifth International FLAIRS Conference.</bibtext> </blist> <blist> <bibtext> Nguyen, D. Q, Vu, T, & Nguyen, A. T. (2020). Bertweet: A pre-trained language model for English tweets. arXiv preprint arXiv:2005.10200.</bibtext> </blist> <blist> <bibtext> Nickow, A, Oreopoulos, P, & Quan, V. (2024). The promise of tutoring for prek–12 learning: A systematic review and meta-analysis of the experimental evidence. American Educational Research Journal, 61(1), 74–107. https://doi.org/10.3102/00028312231208687</bibtext> </blist> <blist> <bibtext> Omarchevska Y, Lachner A, Richter J, Scheiter K. It takes two to tango: How scientific reasoning and self-regulation processes impact argumentation quality. Journal of the Learning Sciences. 2022; 31; 2: 237-277. 10.1080/10508406.2021.1966633</bibtext> </blist> <blist> <bibtext> Ouyang F, Tang Z, Cheng M, Chen Z. Using an integrated discourse analysis approach to analyze a group's collaborative argumentation. Thinking Skills and Creativity. 2023; 47: 101227. 10.1016/j.tsc.2022.101227</bibtext> </blist> <blist> <bibtext> Reichenberg R, Avissar G, Sagee R. 'I owe to my tutor much of my professional development': Looking at the benefits of tutoring as perceived by the tutees. Professional Development in Education. 2015; 41; 1: 40-56. 10.1080/19415257.2013.866974</bibtext> </blist> <blist> <bibtext> Resnick, L. B, Asterhan, C. S, Clarke, S. N, & Schantz, F. (2018). Next generation research in dialogic learning. In G. E. Hall, L. F. Quinn, & D.M. Gollnick (Eds.), Wiley handbook of teaching and learning (pp. 323–338). John Wiley & Sons, Inc.</bibtext> </blist> <blist> <bibtext> Robinette RL. Understanding and meeting the professional development needs of English educators at a private tutoring organization in China. Teaching and Teacher Education. 2024; 138: 104403. 10.1016/j.tate.2023.104403</bibtext> </blist> <blist> <bibtext> Shaffer D, Ruis A. Epistemic network analysis: A worked example of theory-based learning analytics. Handbook of learning analytics. 2017; Society for Learning Analytics Research (SoLAR)</bibtext> </blist> <blist> <bibtext> Shaffer DW. Quantitative ethnography. 2017; Cathcart Press</bibtext> </blist> <blist> <bibtext> Shaffer D, Ruis A. A tutorial on epistemic network analysis: Analyzing the structure of connections in cognitive, social, and interaction data. Journal of Learning Analytics. 2016; 3; 3: 9-45. 10.18608/jla.2016.33.3</bibtext> </blist> <blist> <bibtext> Shan, D, Bhattacharya, P, Kao, B, & Phan, T. (2021). Characterizing amateur tutoring behavior on a large online learning platform. In Proceedings of the Eighth ACM Conference on Learning@ Scale (pp. 239–242).</bibtext> </blist> <blist> <bibtext> Snapask. (2024). A school squeezed into snapask. Retrieved February 1, 2024, from https://snapask.com/</bibtext> </blist> <blist> <bibtext> Song, Y, Hao, T, Liu, Z, & Lan, Z. (2020). A systematic review of frameworks for coding towards classroom dialogue. Emerging Technologies for Education: 4th International Symposium, SETE 2019, Held in Conjunction with ICWL 2019, Magdeburg, Germany, September 23–25, 2019, Revised Selected Papers 4, (pp. 226–236).</bibtext> </blist> <blist> <bibtext> Song Y, Cheng B, Zhu J, Hu X. Exploring the collective process of classroom dialogue using sequential pattern mining technique. International Journal of Educational Research. 2022; 115: 102050. 10.1016/j.ijer.2022.102050</bibtext> </blist> <blist> <bibtext> Št'astny' V. Private tutoring lessons supply: Insights from online advertising in the Czech Republic. Compare: A Journal of Comparative and International Education. 2017; 47; 4: 561-579. 10.1080/03057925.2016.1259064</bibtext> </blist> <blist> <bibtext> Tao Y, Chen G. Coding schemes and analytic indicators for dialogic teaching: A systematic review of the literature. Learning, Culture and Social Interaction. 2023; 39: 100702. 10.1016/j.lcsi.2023.100702</bibtext> </blist> <blist> <bibtext> Teakel, S, Linden, K, van der Ploeg, N, & Roman, N. (2023). Embedding equity: Online tutor support to provide effective feedforward on assessments. Assessment & Evaluation in Higher Education, 49(3), 320–333. https://doi.org/10.1080/02602938.2023.2232955</bibtext> </blist> <blist> <bibtext> Tlili A, Wang H, Gao B, Shi Y, Zhiying N, Looi CK, Huang R. Impact of cultural diversity on students' learning behavioral patterns in open and online courses: A lag sequential analysis approach. Interactive Learning Environments. 2023; 31; 6: 3951-3970. 10.1080/10494820.2021.1946565</bibtext> </blist> <blist> <bibtext> Trygstad PJ, Malzahn KA, Banilower ER, Plumley CL, Bruce AD. Are all students getting equal access to high-quality science education? Data from the 2018 NSSME+. 2020; Horizon Research, Inc</bibtext> </blist> <blist> <bibtext> Tsuei M. Learning behaviours of low-achieving children's mathematics learning in using of helping tools in a synchronous peer-tutoring system. Interactive Learning Environments. 2017; 25; 2: 147-161. 10.1080/10494820.2016.1276078</bibtext> </blist> <blist> <bibtext> Tu, C. H, & McIsaac, M. (2002). The relationship of social presence and interaction in online classes. American Journal of Distance Education, 16(3), 131–150. https://doi.org/10.1207/S15389286AJDE1603_2</bibtext> </blist> <blist> <bibtext> Vail, A. K, Grafsgaard, J. F, Boyer, K. E, Wiebe, E. N, & Lester, J. C. (2016). Predicting learning from student affective response to tutor questions. Intelligent Tutoring Systems: 13th International Conference, ITS 2016, Zagreb, Croatia, June 7–10, 2016. Proceedings 13, 154–164.</bibtext> </blist> <blist> <bibtext> Vail AK, Boyer KE. Identifying effective moves in tutoring: On the refinement of dialogue act annotation schemes. 2014: International conference on intelligent tutoring systems; Springer International Publishing: 199-209</bibtext> </blist> <blist> <bibtext> Vista A, Awwal N, Care E. Sequential actions as markers of behavioural and cognitive processes: Extracting empirical pathways from data streams of complex tasks. Computers & Education. 2016; 92: 15-36. 10.1016/j.compedu.2015.10.009</bibtext> </blist> <blist> <bibtext> Vrikki M, Kershner R, Calcagni E, Hennessy S, Lee L, Hernández F, Estrada N, Ahmed F. The teacher scheme for educational dialogue analysis (t-seda): Developing a research-based observation tool for supporting teacher inquiry into pupils' participation in classroom dialogue. International Journal of Research & Method in Education. 2019; 42; 2: 185-203. 10.1080/1743727X.2018.1467890</bibtext> </blist> <blist> <bibtext> Wang D, Tao Y, Chen G. Artificial intelligence in classroom discourse: A systematic review of the past decade. International Journal of Educational Research. 2024; 123: 102275. 10.1016/j.ijer.2023.102275</bibtext> </blist> <blist> <bibtext> Wood D, Bruner JS, Ross G. The role of tutoring in problem solving. Journal of Child Psychology and Psychiatry. 1976; 17; 2: 89-100. 10.1111/j.1469-7610.1976.tb00381.x</bibtext> </blist> <blist> <bibtext> Wu SY, Chen SY, Hou HT. A study of users' reactions to a mixed online discussion model: A lag sequential analysis approach. International Journal of Human-Computer Interaction. 2015; 31; 3: 180-192. 10.1080/10447318.2014.986637</bibtext> </blist> <blist> <bibtext> Xu, S, Ding, W, & Liu, Z. (2020). Automatic dialogic instruction detection for k-12 online one-on-one classes. Artificial Intelligence in Education: 21st International Conference, AIED 2020, Ifrane, Morocco, July 6–10, 2020, Proceedings, Part II 21, 340– 345.</bibtext> </blist> <blist> <bibtext> Zhang L, Pan M, Yu S, Chen L, Zhang J. Evaluation of a student-centered online one-to-one tutoring system. Interactive Learning Environments. 2023; 31; 7: 4251-4269. 10.1080/10494820.2021.1958234</bibtext> </blist> <blist> <bibtext> Zhang S, Chen J, Wen Y, Chen H, Gao Q, Wang Q. Capturing regulatory patterns in online collaborative learning: A network analytic approach. International Journal of Computer-Supported Collaborative Learning. 2021; 16: 37-66. 10.1007/s11412-021-09339-5</bibtext> </blist> </ref> <aug> <p>By Deliang Wang; Lei Gao; Dapeng Shan; Gaowei Chen; Chenwei Zhang and Ben Kao</p> <p>Reported by Author; Author; Author; Author; Author; Author</p> </aug> <nolink nlid="nl1" bibid="bib20" firstref="ref1"></nolink> <nolink nlid="nl2" bibid="bib25" firstref="ref2"></nolink> <nolink nlid="nl3" bibid="bib74" firstref="ref3"></nolink> <nolink nlid="nl4" bibid="bib12" firstref="ref5"></nolink> <nolink nlid="nl5" bibid="bib17" firstref="ref6"></nolink> <nolink nlid="nl6" bibid="bib26" firstref="ref7"></nolink> <nolink nlid="nl7" bibid="bib31" firstref="ref8"></nolink> <nolink nlid="nl8" bibid="bib37" firstref="ref9"></nolink> <nolink nlid="nl9" bibid="bib49" firstref="ref10"></nolink> <nolink nlid="nl10" bibid="bib54" firstref="ref11"></nolink> <nolink nlid="nl11" bibid="bib62" firstref="ref12"></nolink> <nolink nlid="nl12" bibid="bib77" firstref="ref13"></nolink> <nolink nlid="nl13" bibid="bib59" firstref="ref15"></nolink> <nolink nlid="nl14" bibid="bib64" firstref="ref16"></nolink> <nolink nlid="nl15" bibid="bib67" firstref="ref19"></nolink> <nolink nlid="nl16" bibid="bib23" firstref="ref21"></nolink> <nolink nlid="nl17" bibid="bib33" firstref="ref22"></nolink> <nolink nlid="nl18" bibid="bib34" firstref="ref27"></nolink> <nolink nlid="nl19" bibid="bib42" firstref="ref30"></nolink> <nolink nlid="nl20" bibid="bib32" firstref="ref31"></nolink> <nolink nlid="nl21" bibid="bib24" firstref="ref36"></nolink> <nolink nlid="nl22" bibid="bib43" firstref="ref37"></nolink> <nolink nlid="nl23" bibid="bib29" firstref="ref38"></nolink> <nolink nlid="nl24" bibid="bib35" firstref="ref39"></nolink> <nolink nlid="nl25" bibid="bib73" firstref="ref40"></nolink> <nolink nlid="nl26" bibid="bib53" firstref="ref42"></nolink> <nolink nlid="nl27" bibid="bib44" firstref="ref43"></nolink> <nolink nlid="nl28" bibid="bib10" firstref="ref45"></nolink> <nolink nlid="nl29" bibid="bib46" firstref="ref46"></nolink> <nolink nlid="nl30" bibid="bib27" firstref="ref47"></nolink> <nolink nlid="nl31" bibid="bib60" firstref="ref48"></nolink> <nolink nlid="nl32" bibid="bib63" firstref="ref49"></nolink> <nolink nlid="nl33" bibid="bib45" firstref="ref50"></nolink> <nolink nlid="nl34" bibid="bib28" firstref="ref52"></nolink> <nolink nlid="nl35" bibid="bib72" firstref="ref55"></nolink> <nolink nlid="nl36" bibid="bib47" firstref="ref56"></nolink> <nolink nlid="nl37" bibid="bib70" firstref="ref57"></nolink> <nolink nlid="nl38" bibid="bib36" firstref="ref58"></nolink> <nolink nlid="nl39" bibid="bib69" firstref="ref60"></nolink> <nolink nlid="nl40" bibid="bib38" firstref="ref61"></nolink> <nolink nlid="nl41" bibid="bib76" firstref="ref65"></nolink> <nolink nlid="nl42" bibid="bib58" firstref="ref70"></nolink> <nolink nlid="nl43" bibid="bib56" firstref="ref72"></nolink> <nolink nlid="nl44" bibid="bib78" firstref="ref73"></nolink> <nolink nlid="nl45" bibid="bib55" firstref="ref74"></nolink> <nolink nlid="nl46" bibid="bib57" firstref="ref75"></nolink> <nolink nlid="nl47" bibid="bib50" firstref="ref76"></nolink> <nolink nlid="nl48" bibid="bib51" firstref="ref77"></nolink> <nolink nlid="nl49" bibid="bib16" firstref="ref80"></nolink> <nolink nlid="nl50" bibid="bib61" firstref="ref81"></nolink> <nolink nlid="nl51" bibid="bib71" firstref="ref82"></nolink> <nolink nlid="nl52" bibid="bib65" firstref="ref84"></nolink> <nolink nlid="nl53" bibid="bib40" firstref="ref86"></nolink> <nolink nlid="nl54" bibid="bib75" firstref="ref87"></nolink> <nolink nlid="nl55" bibid="bib18" firstref="ref88"></nolink> <nolink nlid="nl56" bibid="bib13" firstref="ref91"></nolink> <nolink nlid="nl57" bibid="bib39" firstref="ref92"></nolink> <nolink nlid="nl58" bibid="bib48" firstref="ref93"></nolink> <nolink nlid="nl59" bibid="bib14" firstref="ref94"></nolink> <nolink nlid="nl60" bibid="bib21" firstref="ref96"></nolink> <nolink nlid="nl61" bibid="bib41" firstref="ref97"></nolink> <nolink nlid="nl62" bibid="bib52" firstref="ref100"></nolink> <nolink nlid="nl63" bibid="bib68" firstref="ref101"></nolink> <nolink nlid="nl64" bibid="bib30" firstref="ref102"></nolink> <nolink nlid="nl65" bibid="bib19" firstref="ref105"></nolink> <nolink nlid="nl66" bibid="bib11" firstref="ref106"></nolink> <nolink nlid="nl67" bibid="bib15" firstref="ref107"></nolink> <nolink nlid="nl68" bibid="bib66" firstref="ref109"></nolink> <nolink nlid="nl69" bibid="bib22" firstref="ref110"></nolink>
Header DbId: eric
DbLabel: ERIC
An: EJ1497295
AccessLevel: 3
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: When Tutors Simultaneously Instruct Students from the Primary, Middle, and High School Levels in Online One-on-One Tutoring: Investigating the Interaction Dynamics Using AI, ENA, and LSA Methods
– Name: Language
  Label: Language
  Group: Lang
  Data: English
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Deliang+Wang%22">Deliang Wang</searchLink> (ORCID <externalLink term="http://orcid.org/0009-0008-6488-0234">0009-0008-6488-0234</externalLink>)<br /><searchLink fieldCode="AR" term="%22Lei+Gao%22">Lei Gao</searchLink><br /><searchLink fieldCode="AR" term="%22Dapeng+Shan%22">Dapeng Shan</searchLink><br /><searchLink fieldCode="AR" term="%22Gaowei+Chen%22">Gaowei Chen</searchLink><br /><searchLink fieldCode="AR" term="%22Chenwei+Zhang%22">Chenwei Zhang</searchLink><br /><searchLink fieldCode="AR" term="%22Ben+Kao%22">Ben Kao</searchLink>
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="SO" term="%22Journal+of+Science+Education+and+Technology%22"><i>Journal of Science Education and Technology</i></searchLink>. 2025 34(5):1128-1142.
– Name: Avail
  Label: Availability
  Group: Avail
  Data: Springer. Available from: Springer Nature. One New York Plaza, Suite 4600, New York, NY 10004. Tel: 800-777-4643; Tel: 212-460-1500; Fax: 212-460-1700; e-mail: customerservice@springernature.com; Web site: https://link.springer.com/
– Name: PeerReviewed
  Label: Peer Reviewed
  Group: SrcInfo
  Data: Y
– Name: Pages
  Label: Page Count
  Group: Src
  Data: 15
– Name: DatePubCY
  Label: Publication Date
  Group: Date
  Data: 2025
– Name: TypeDocument
  Label: Document Type
  Group: TypDoc
  Data: Journal Articles<br />Reports - Research
– Name: Audience
  Label: Education Level
  Group: Audnce
  Data: <searchLink fieldCode="EL" term="%22Elementary+Education%22">Elementary Education</searchLink><br /><searchLink fieldCode="EL" term="%22Junior+High+Schools%22">Junior High Schools</searchLink><br /><searchLink fieldCode="EL" term="%22Middle+Schools%22">Middle Schools</searchLink><br /><searchLink fieldCode="EL" term="%22Secondary+Education%22">Secondary Education</searchLink><br /><searchLink fieldCode="EL" term="%22High+Schools%22">High Schools</searchLink>
– Name: Subject
  Label: Descriptors
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22Tutors%22">Tutors</searchLink><br /><searchLink fieldCode="DE" term="%22Tutoring%22">Tutoring</searchLink><br /><searchLink fieldCode="DE" term="%22Peer+Teaching%22">Peer Teaching</searchLink><br /><searchLink fieldCode="DE" term="%22Computer+Mediated+Communication%22">Computer Mediated Communication</searchLink><br /><searchLink fieldCode="DE" term="%22Interaction%22">Interaction</searchLink><br /><searchLink fieldCode="DE" term="%22Elementary+School+Students%22">Elementary School Students</searchLink><br /><searchLink fieldCode="DE" term="%22Middle+School+Students%22">Middle School Students</searchLink><br /><searchLink fieldCode="DE" term="%22High+School+Students%22">High School Students</searchLink><br /><searchLink fieldCode="DE" term="%22Mathematics+Instruction%22">Mathematics Instruction</searchLink><br /><searchLink fieldCode="DE" term="%22Teaching+Methods%22">Teaching Methods</searchLink><br /><searchLink fieldCode="DE" term="%22Tutor+Training%22">Tutor Training</searchLink><br /><searchLink fieldCode="DE" term="%22Teaching+Skills%22">Teaching Skills</searchLink><br /><searchLink fieldCode="DE" term="%22Dialogs+%28Language%29%22">Dialogs (Language)</searchLink><br /><searchLink fieldCode="DE" term="%22Instructional+Effectiveness%22">Instructional Effectiveness</searchLink><br /><searchLink fieldCode="DE" term="%22Efficiency%22">Efficiency</searchLink>
– Name: DOI
  Label: DOI
  Group: ID
  Data: 10.1007/s10956-024-10154-4
– Name: ISSN
  Label: ISSN
  Group: ISSN
  Data: 1059-0145<br />1573-1839
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: Online one-on-one tutoring serves as a personalized approach to supplement classroom instruction. However, with the growing tutoring market, a single tutor often handles inquiries from students across primary, middle, and high school levels. Consequently, the extent of tutors' interactions with students of varying grades and their use of tutoring strategies to enhance student learning remains unclear. To address this gap, we collected and analyzed 1500 tutoring dialogues from amateur mathematics tutors concurrently instructing students from primary, middle, and high school levels. These dialogues were annotated using a coding scheme and a well-trained powerful artificial intelligence (AI) model. The interaction dynamics were subsequently examined using epistemic network analysis and lag sequential analysis, yielding findings on the occurrences, co-occurrences, and sequential patterns of dialogic strategies. First, the results reveal that tutors frequently engaged in off-topic behaviors and direct instruction, regardless of students' educational level. Second, tutors' constructive questions and knowledge sharing and instruction were more associated with greater constructive expressions from students at higher educational levels, while primary students primarily demonstrated simple acknowledgment. Third, tutors exhibited limited sequential patterns of dialogic strategies when tutoring primary and middle school students, mainly focusing on question-asking behaviors and evaluation and feedback. In contrast, tutors displayed diverse patterns across various categories of dialogic strategies when instructing high school students, emphasizing the facilitation of students' reasoning and metacognition. These findings underscore the importance of training tutors to develop dialogic skills and adopt tailored pedagogical approaches for different educational levels, ensuring effective and efficient online one-on-one mathematics tutoring.
– Name: AbstractInfo
  Label: Abstractor
  Group: Ab
  Data: As Provided
– Name: DateEntry
  Label: Entry Date
  Group: Date
  Data: 2026
– Name: AN
  Label: Accession Number
  Group: ID
  Data: EJ1497295
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=eric&AN=EJ1497295
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1007/s10956-024-10154-4
    Languages:
      – Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 15
        StartPage: 1128
    Subjects:
      – SubjectFull: Tutors
        Type: general
      – SubjectFull: Tutoring
        Type: general
      – SubjectFull: Peer Teaching
        Type: general
      – SubjectFull: Computer Mediated Communication
        Type: general
      – SubjectFull: Interaction
        Type: general
      – SubjectFull: Elementary School Students
        Type: general
      – SubjectFull: Middle School Students
        Type: general
      – SubjectFull: High School Students
        Type: general
      – SubjectFull: Mathematics Instruction
        Type: general
      – SubjectFull: Teaching Methods
        Type: general
      – SubjectFull: Tutor Training
        Type: general
      – SubjectFull: Teaching Skills
        Type: general
      – SubjectFull: Dialogs (Language)
        Type: general
      – SubjectFull: Instructional Effectiveness
        Type: general
      – SubjectFull: Efficiency
        Type: general
    Titles:
      – TitleFull: When Tutors Simultaneously Instruct Students from the Primary, Middle, and High School Levels in Online One-on-One Tutoring: Investigating the Interaction Dynamics Using AI, ENA, and LSA Methods
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Deliang Wang
      – PersonEntity:
          Name:
            NameFull: Lei Gao
      – PersonEntity:
          Name:
            NameFull: Dapeng Shan
      – PersonEntity:
          Name:
            NameFull: Gaowei Chen
      – PersonEntity:
          Name:
            NameFull: Chenwei Zhang
      – PersonEntity:
          Name:
            NameFull: Ben Kao
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 01
              Type: published
              Y: 2025
          Identifiers:
            – Type: issn-print
              Value: 1059-0145
            – Type: issn-electronic
              Value: 1573-1839
          Numbering:
            – Type: volume
              Value: 34
            – Type: issue
              Value: 5
          Titles:
            – TitleFull: Journal of Science Education and Technology
              Type: main
ResultId 1